You are on page 1of 24

Lessons on Writing Papers

Stanley Bak
1/23/2009

History

Paper submitted in 2008 or !"SS

!e#e$ted

Paper submitted in 2009 to !"%S

%$$epted

"&e dieren$e'

(ain )ontribution

Simple* %r$&ite$ture on an +P,%

We mo-e t&e Saety Subsystem and .e$ision


Logi$ to &ard/are or additional saety

)&anges

(oti-ation or $&anges

!e-ie/er suggestions

The Craft of Scientific Writing 0 %lley

)ombined /it& re$ommendations rom


proessors1 o $ourse

!e-ie/er Suggestions

!e-ie/er )omments

Se-eral $ategories o $omments

(isunderstandings 2re/ord and $lariy3

Small )&anges 2modiy senten$e or


paragrap&3

Large )&anges 2add/remo-e se$tions3



(isunderstandings

4t&ere is a serious problem /&en t&e de$ision


logi$ does not re$ei-e signals rom t&e $omple*
pro$essor555 "&e paper #ust states t&at t&e last
-alue is stored and used5 6 ask mysel1 &o/ long
is su$& a stored -alid'7

8nne$essary implementation details may


$onuse readers
We store the most-recent output for each of the
controllers and use that value when a decision must
be made

Small )&anges )omments

4W&y +P,% and not sot/are'7

%greed9 :ou $an use sot/are1 i you a$$ept its


asso$iated $omple*ity

4:our e*ample is bad be$ause5557

!emo-e e*ample

4:ou said a ormally -eriied ;S is ineasible but


/&at about t&e /ork by5557

)&ange to 4Although great progress has been made


towards producing a verifiable RTOS[3] most current
RTOSes have been neither formall! verified nor
e"haustivel! tested7

Small )&anges )omments 223

4"&is isn<t ne/ or interesting7

=mp&asi>e /&at is ne/ and non0ob-ious1


and say /&y it isn<t ob-ious
Two Simple" safet!-critical components the safet!
controller and decision module are moved to a
dedicated processing unit not for the typical
HW/SW co-design reasons of power and
performance but instead to provide isolation from
software-related comple"it!#

Small )&anges )omments 233

4"&ere<s no mention o limitations7

?ot mentioning limitations implies t&e solution is


always easier1 $&eaper and better

/&i$& makes re-ie/ers ner-ous


Simple" $both versions% should not be regarded as a
one-si&e-fits-all robustness approach# To guarantee
safet! we must [do ' ( and ) which are not
alwa!s possible]#

Small )&anges )omments 2@3

;t&er mentions o limitations


An inverted pendulum however does not
completel! lend itself to our end-to-end design
process#
One drawbac* of the S!stem-+evel version of
Simple" is that additional resources are re,uired to
run the decision module and comple" controller
such as the -./A#

Large )&anges

;riginally /e &ad a se$tion /&i$& ormally


model0$&e$ked t&e Simple* %r$&ite$ture or
saety gi-en $ertain properties o t&e $ontrollers

"&e re-ie/ers /ere not impressed

4=*$ept or inormal arguments1 t&ere is no proo


t&at t&ese @ are ne$essary and sui$ient to
guarantee t&e $orre$tness o t&e ar$&ite$ture57

Large )&anges 223

6nstead1 t&e re-ie/ers &inted to/ards a more


interesting dire$tion

4"&e problem is t&at t&ere don<t seem to be any


s$ien$e in t&e sense o a systemati$ /ay o iguring
out t&e $orre$t implementation o t&e de$ision logi$
or t&e saety subsystem57

Solution9 add se$tion about modeling $&e$king


t&e de$ision logi$ and saety subsystem

Large )&anges 233

We also added a se$tion about %%.L


generation1 /&i$& generated re-ie/er
$omments

4%ut&ors s&ould gi-e a brie introdu$tion or related


reeren$es on t&e %%.L model57

4"&e %%.L material s&ould be impro-ed and better


moti-ated1 as it is it seems to add little or not&ing to
t&e paper57

4"&e )rat o S$ientii$ Writing7
(i$&ael %lley

,eneral )&anges rom ");SW

%-oid pretentious /ords9

8tili>e / 8tili>ation 0A 8se

+a$ilitate 0A )ase / Bring %bout

)omponent 0A Part

)&ange title9

.emonstrate $lear mat$& to one o t&e


$oneren$e topi$s

4novel hardware or system architectures


for real-time or embedded systems7

)&ange "itle

+rom ");SW1 titles s&ould do t/o t&ings

6dentiy ield o study

Separate rom all ot&er publi$ations in t&at ield

Pre-ious "itle

The Hardware Simplex rchitecture for !eal-


Time" Safety-Critical Systems

?e/ title

The System-#evel Simplex rchitecture for


$mproved !eal-Time %mbedded System Safety

List ormating

Lists s&ould be o 21 31 or @ items

Lists s&ould be $on$ise

Lists s&ould be pre$ise

Parallel stru$ture

(ain )ontributions

original9

i*ed9

Saety !ules 2original3

Long list o se-en design


rules

Spanning t/o
pages

Some /it& multiple


paragrap&s

Saety !ules 2i*ed3

We $&anged t&e saety rules to t&ree


$ategories o ne$essary properties
We have identified several necessar! conditions re,uired
b! a S!stem-+evel Simple" design that are chec*ed b!
our OSAT0 architecture chec*er# The properties can be
classified into resource isolation properties data
consistenc! properties and data flow properties# We
briefl! describe each of these along with associated
failures that ma! occur if the! are not present#

%dded ;t&er Lists

%ll items are


Buestions

%ll Buestions start


/it& dierent /ord

Paper Stru$ture

;riginal9

6ntrodu$tion

Related Work

HW Simple* .esign

)ase Studies

Pa$emaker

6n-erted Pendulum

HW Simplex Model

)on$lusions

!e-ised9

6ntrodu$tion

SL Simple* .esign

Design Process

)ase Studies

Pa$emaker

6n-erted Pendulum

Related Work

)on$lusions

Cuestions D )omments'

6<ll send out bot& -ersions o t&e paper and t&is


presentation555

You might also like